kentoh - Fotolia
Part converged and part hyper-converged, composable infrastructure combines characteristics of both those types of environments with programmatic and disaggregated features and more to evolve IT infrastructure and make it more dynamic. Composable infrastructure vendors promise to deliver in their products more agile IT environments that better meet the business and IT needs of today's fast-paced enterprises.
Composable infrastructure does this by collecting discrete pools -- compute, storage, network fabric, etc. -- of resources and delivers those resources as services. An administrator can then quickly draw up these on-demand IT resources for specific applications, use cases and workloads without needing to configure hardware. The potential for vendor lock-in has been cited as a concern with composable infrastructure, however, due to a lack of standards for interoperability and each vendor approaching the technology from its own angle.
See part one in this two-part glossary series to get up to speed on important terms you will run across as you consider this fledgling -- and rapidly evolving -- disaggregated rack-scale infrastructure. In this second part, we highlight composable infrastructure vendors and products on the market today that put this new data center paradigm into practice.
HPE Synergy. In 2015, Hewlett Packard Enterprise launched HPE Synergy, essentially establishing composable infrastructure as a viable IT environment option to go with traditional, converged and hyper-converged infrastructures. HPE Synergy is a composable blade infrastructure that has a 10U rack-scale server frame, with HPE storage blades and an interconnected network fabric. You can link together multiple chassis through the HPE Synergy Link Module and swap in compute, storage and networking power management modules as required.
HPE Synergy focuses on software control and automation via templates, policies and a universal API. Composer and Image Streamer run through HPE OneView to provide the platform's monitoring, management and orchestration engine. Check out a case study of HPE Synergy in action.
HPE SimpliVity with Composable Fabric. Unlike Synergy, a pure composable infrastructure play, HPE SimpliVity with Composable Fabric is a hyper-converged infrastructure system with composable features. It enables these by using software-defined technologies to manage and automate compute, storage and network resources.
That's because, as a software-defined networking product, Composable Fabric interconnects infrastructure nodes -- e.g., switches and routers -- in a mesh fabric, rather than a traditional hierarchical tree structure. See more about HPE SimpliVity with Composable Fabric.
Cisco UCS. Around for years, Cisco's Unified Computing System (UCS) is a converged infrastructure system that is, perhaps, one of the first examples of a composable infrastructure -- although it hasn't been sold or promoted as such. Hardware features include storage servers, switches, and rack and blade servers, while the UCS Manager software serves as a composing engine by using Service Profiles to generate bare-metal compute instances and storage configurations at the storage device level. Read more on Cisco UCS.
Dell EMC PowerEdge MX7000. Introduced in 2018, Dell EMC's PowerEdge MX7000 is composable infrastructure in all but name. Dell EMC bills the MX7000 as "kinetic infrastructure" due to its bridging of traditional and software-defined data center technologies through the housing of disaggregated blocks of compute, storage and networking in a 7U modular chassis running 14th-generation PowerEdge servers.
Users can scale a fabric up to 10 PowerEdge MX7000 chassis and can quickly assign shared pools of resources to specific workloads on the fly, and then return those resources to the shared pool when no longer needed. Dell provides integrated management through Dell's OpenManage software suite, and the environment is integrated with BMC, Red Hat Ansible, Microsoft System Center and VMware vCenter. Learn more about the PowerEdge MX7000.
Liqid PCIe fabric switch. Liqid is a startup among composable infrastructure vendors. Its Grid PCIe 3.0 is a PCIe Gen 3 internal fabric switch that, along with the company's composable software, lets users disaggregate and pool compute, flash, network interface card and GPU resources using standard bare-metal servers.
Targeted at GPU-based application and high-performance use cases such as AI, Liqid's Command Center Software serves as the composing engine that does the disaggregating of hardware resources within the servers and assigns them as virtual servers to specific workloads.
In April, Liqid announced it would provide unified multifabric support for its composable infrastructure product by adding PCIe Gen 4, Ethernet and InfiniBand -- along with laying a foundation for the upcoming Gen-Z specifications -- in Command Center 2.2, which is due later this year. See more on Liqid composability.
DriveScale. Established in 2016, DriveScale's software maps existing diskless servers (compute) and storage enclosures across a fabric via Ethernet, so these resources can be pooled into adaptable servers that can be assigned to an application or workload. Server agents on each compute server collect status and configuration data, while a composer software creates compute instances that run on virtual machine instances or bare-metal Linux.
Users manage multiple DriveScale deployments and locations via a software-as-a-service-based portal, called Cloud Central. When it comes to expansion, DriveScale enables users to purchase additional compute and storage separately. Learn more about DriveScale.
Kaminario Flex. Software to enable composability in a Kaminario all-flash environment -- for the past year, Kaminario arrays have been available from resellers as a reference design -- Kaminario Flex enables users to compose, manage and free up storage resources for workloads as they are created, moved or changed without having to physically move or recable disks. Flex's orchestration and automation tools are available through a dashboard that's compatible with Kaminario VisionOS and Kaminario Clarity.
The dashboard lets users create new virtual storage instances or virtual private arrays, similar to the way other composable infrastructure products we've covered enable users to instantiate virtual servers for specific workloads. Find out more about Kaminario Flex and all-flash arrays.
Western Digital OpenFlex. Western Digital bills OpenFlex composable infrastructure as a flexible way to avoid overprovisioning and underutilization of storage and compute resources. OpenFlex uses a common interface and API to disaggregate and pool all compute and storage capacity to allow for independent scaling of those resources.
Hardware consists of Western Digital's F3000 Series NVMe array, which scales up to 61 TB of NAND flash; D3000 disk array, with 168 TB of raw capacity; and E3000 fabric enclosure. It uses NVMe-oF protocols to address all disk and flash storage devices. Read more about Western Digital's composable OpenFlex platform.
Lenovo TruScale. This is not your usual composable infrastructure product -- if there even is such a thing. Nor does Lenovo advertise it as composable. But Lenovo TruScale has composable features that enable users to purchase the company's storage hardware and its ThinkSystem compute servers together as a scalable service.
Lenovo provides remote management and installs the hardware in the customer's data center without requiring a minimum capital purchase, with enterprises only having to pay for the resources -- compute and storage -- they consume. Learn more about Lenovo TruScale.
Datrium DVX: Datrium bills itself as a hyper-converged infrastructure vendor, although it offers separate compute and storage -- disk or flash -- nodes through split provisioning with its DVX platform. All workloads run locally on the compute nodes, which also run software with features such as compression, replication, encryption, deduplication and snapshots to manage and protect data across the storage nodes.
Like products from those that bill themselves as composable infrastructure vendors, the platform disaggregates storage and makes it available as a single composable pool for workloads. It scales up to 128 compute hosts and more than a petabyte in capacity over 10 data nodes. Read more on Datrium DVX.
Intel Rack Scale Design. Unique among composable infrastructure vendors, Intel's Rack Scale Design is a disaggregated, composable reference architecture that aims to deliver hyperscalability to more traditional enterprise environments using top-of-rack switches, storage and rack-based servers. In Intel Rack Scale Design, one or more racks comprise a pod, which you can think of as being equivalent to an HCI node.
Through a web-based user interface, the platform's Pod Manager software and Distributed Management Task Force Redfish APIs enable communication between racks in pods and allow users to compose bare-metal compute environments for specific workloads. Intel Rack Scale Design offerings are available from Gigabyte, Supermicro, Dell EMC, Ericsson, American Megatrends, QCT, Inspur and others. Learn more about the Intel Rack Scale Design.